Description of problem: We found that if we run kubevirt test with -installed-namespace=openshift-cnv the CDI deployment is deleted. With installed-namespace=kubevirt-hyperconverged all works. Version-Release number of selected component (if applicable): CNV-2.2 It's tests BZ.
I wonder if we're about to make a similar bug report for how it manages storageclasses (unconditionally creates at the start of testing, and destroys at the end) https://github.com/kubevirt/kubevirt/blob/d072d7ffcbba8d862e79db7749fd0effe2ca2f4e/tests/utils.go#L556-L559
Targeting 2.3.0 because we can avoid this issue by not using --deploy-testing-infra. This flag sets up missing bits in the environment, our test environment should have all the parts it needs to run the tests.
Minimal risk for 2.2 release so let's try to backport it there
I initially assumed that we want the flag to not exist because it's likely to damage clusters. That's a really risky thing to backport at all. What are you hoping to fix?
We just want to make sure the 2.2 branch of kubevirt functional tests has your fix too so that the tests don't cause this issue in the future.
Still confused. I believe we should avoid using this code by not passing the flag --deploy-testing-infra Downstream always has CDI. We don't need the kubevirt tier 1 tests to deploy CDI for us. The flag doesn't do anything positive, and it does do something negative (delete CDI in the end).
Tareq can we remove the command line flag '--deploy-testing-infra' from the d/s test invocation since all components will have been installed already by HCO?
Tareq says he would like to keep using this flag. I don't think there's anything to do in this bug, unless I misunderstood it.
Israel is there anything to do or has this now been resolved?
We can close this BZ.